Industrial Organization I

1
Industrial Organization I

• Econometric Identification

2
The Problem

• Empirical economists mainly concerned with
  identification
• Understanding what is the causal relationship
  between variables
• This is perhaps the MOST important issue in
  empirical work
• Correlations are a good place to start but not a
  good place end.

3
The Problem

• Identification problems come in different
  flavors
• Structural models Example is identifying demand
  slope. Would regressing quantity on price tell us
  something reliable?
• Policy questions
• Does increasing number of policemen decrease
  crime?
• Does the Fed interest rate affect GDP?
• Management questions
• Does the type of management affect productivity?

4
Does this mean demand slopes up?

Q
5
(No Transcript)
6
Reasons for Correlation

• Lets assume variables Yt and Xt are correlated
• There can be three reasons for this
• Changes in Yt drive changes in Xt
• Changes in Xt drive changes in Yt
• Correlated variable Changes in Zt drives Xt and
  Yt
• How do we make sure we are measuring what we want?

7
Ways to Achieve Identification

• Experiments you generate the variation
• Natural Experiments you know what generated the
  variation
• Instrumental variables you hope you can get a
  variable that can provide you variation

8
Experiments

• Experiments are standard in Science Medicine
• For example
• Set up a treatment and control group for a new
  drug, making sure these are comparable (or
  randomly selected)
• Ensure the sample sizes are large enough to
  obtain statistical significance
• Ensure there are no confounding effects i.e.
  placebo and treatment groups are as similar as
  possible
• Run the experiment and compare the differential
  effects.

9
Experiments

• Hypothetical example (1)
• Monopolist knows demand will not move for 1
  month
• Changes prices every day
• Records the quantity sold at each price
• Unrealistic example but doable (i.e. controlling
  for trends, seasonalities, etc.)

10
Experiments

• Hypothetical example (2)
• Choose two similar states (e.g. North Carolina
  and South Carolina)
• Increase police force in North, keep the same in
  South
• Look at how much crime decreased in North, after
  accounting for common crime trends (provided by
  control group South)

11
Experiments (Lab and Field)

• Experiments are rare in economics, they are
  expensive. Although they are becoming more
  popular
• Development economics cheaper to run
  experiments in the third World
• Consumer economics small stakes experiments
  that are easy to administer
• Individual business applications firms can
  finance these

12
Natural Experiments

• Fortunate situations that allow identification
• Essence is similar to experiments, but not as
  ideal
• There might be confounding effects because the
  experiment was not designed for our purposes
• Example Beer tax
• In January 1991, beer tax doubled from 9 to
  18/ barrel, increasing all brewers marginal
  cost by the same amount
• Unfortunately, there is no control group, but
  you can study firm behavior (for example)
• How did firms prices react to this change?
• How would firms react to this change under
  different competition models?

13
Natural Experiments

• Example Want to understand the impact of small
  firms employment tax credit
• Assume credit introduced in 2000 for firms with
  250 or less employees
• So could look at firms before and after credit
• But other things also changing (2000 peak of
  dotcom boom, etc.)
• So need to set up a control group of companies
  look similar to firms getting the credit except
  dont get the credit
• Compare firms with 240 employees to those with
  260
• Compare differences
• Between pre and post the credit (1999 versus
  2001)
• Between the treated (240 employees) and
  untreated firms (260 employees)

14
Instrumental Variables

• Assume estimating equation below in Ordinary
  Least Squares
• Y ßX e
• The estimate of ß (XX)-1XY
• (XX)-1X (ßX e)
• E(ß) ß E(eX)/E(XX)
• ß only if e and X are
  uncorrelated
• But if e and X are correlated then the estimated
  is biased, and X is called endogenous
  (correlated with the error)

15
Instrumental Variables

• Thus, estimation of the following demand equation
  would be biased
• Qi a b1 Pi ei
• since Pi and ei are correlated because of
  simultaneity (one can prove that Pi and ei are
  correlated)
• Eb1 ? b1
• Because we ignore simultaneity (supply and demand
  usually move together), this makes the error term
  correlated with price and we get a biased
  estimate of the demand slope.

16
Instrumental Variables

• Imagine we had a variable called an instrument
  Z that was correlated with price but not with
  the demand error term (vi).
• Basically, what we want is to come up with a
  variable that gives us the equivalent of an
  experiment
• In the case of demand and supply, we can use
  supply shifters
• Cost shock such as weather

17
Instrumental Variables

• In the case of schooling and earnings (where
  ability is unobserved), one could use a variable
  such as born on Tuesday (Government told
  everyone born on Tuesday to spend 1 more year in
  school)

18
Instrumental Variables

• In practice instruments are often hard to find.
• IVs are central to empirical work that we will
  see in this class.
• No good instruments unreliable inference.

19
Instrumental Variables

• In practice instruments are often hard to find.
• IVs are central to empirical work that we will
  see in this class.
• No good instruments unreliable inference.